Cleaned and improved

app.py

@@ -7,10 +7,9 @@ from typing import Any, Dict
 from fastapi import FastAPI, Request
 from fastapi.responses import JSONResponse, StreamingResponse
 
-from agent.llm import DeepSeekClient
+from agent.llm.deepseek import DeepSeekClient
 from agent.memory import Memory
 from agent.agent import Agent
-from agent.commands import CommandRegistry
 
 app = FastAPI(
     title="LibreChat Agent Backend",
@@ -20,7 +19,6 @@ app = FastAPI(
 llm = DeepSeekClient()
 memory = Memory()
 agent = Agent(llm=llm, memory=memory)
-commands_registry = CommandRegistry(memory=memory)
 
 
 def extract_last_user_content(messages: list[Dict[str, Any]]) -> str:
@@ -42,12 +40,8 @@ async def chat_completions(request: Request):
     user_input = extract_last_user_content(messages)
     print("Received chat completion request, stream =", stream, "input:", user_input)
 
-    # 🔹 1) Si c'est une commande, on ne fait PAS intervenir le LLM
-    if user_input.strip().startswith("/"):
-        answer = commands_registry.handle(user_input)
-    else:
-        # 🔹 2) Sinon, logique agent + LLM comme avant
-        answer = agent.step(user_input)
+    # Process user input through the agent
+    answer = agent.step(user_input)
 
     # Ensuite = même logique de réponse (non-stream ou stream)
     created_ts = int(time.time())